Introduction

Wildland fuels have accumulated in many western forests of the United States (U.S.) for at least the past 70 years due to 20^th century settlement and management activities (Agee 1998, Hessburg and Agee 2003), and, to some extent, changing climatic conditions (Burkett and others 2005, Schoennagel and others 2004). As demonstrated by recent wildland fires, added fuels are fostering more intense wildfires that are more difficult to contain and control. Consequently, valuable property and natural resources have been destroyed, costs of fire management have escalated, fire-dependent forest ecosystems have deteriorated, and risks to human life and property continue to escalate (GAO 2002, 2003, 2004).

Historically, fires of varying size, frequency, and intensity maintained spatial patterns of forest vegetation, as well as temporal variation in those patterns (Agee 2003, Hessburg and others 2005, Schoennagel and others 2004, Turner 1989). In fact, many agents interacted to shape vegetation patterns and their spatio-temporal variation, including forest insect outbreaks, forest diseases, fires, weather and climatic events, and intentional aboriginal burning (Hessburg and others 2005, Whitlock and Knox 2002). Their interactions resulted in characteristic landscape patterns and caused variation in forest structural attributes, species composition, and habitats that resonated with the dominant disturbance processes. Patterns of forest vegetation were directly linked with the processes that created and maintained them (Hessburg and others 2005, Pickett and White 1985, Turner and others 2001).

Circumstances are quite different today. Patterns and processes are still tightly linked, but not as before. Human influences have created anomalous vegetation patterns, and these patterns support fire, insect, and disease processes that display uncharacteristic duration, spatial extent, and intensity (Ferry and others 1995, Hessburg and others 2005, Kolb and others 1998). For example, 20^th century fire suppression and prevention programs significantly reduced fire frequency in many dry mixed coniferous forests. Contemporary wildland fires are now larger and more intense on average than those of the prior 2 or 3 centuries (GAO 2002, 2003, 2004, U.S. Government 2003, and references therein). In short, settlement and management activities have altered spatial patterns of forest structure, composition, snags, and down wood at patch to province scales. As a result, significant changes in fire frequency, severity, and spatial extent are linked to changes in forest vegetation patterns at patch to province scales (Agee 1998, 2003, Ferry and others 1995, Hessburg and Agee 2003).

Here, we present a decision support system for evaluating wildland fire danger and prioritizing subwatersheds for vegetation and fuels treatment. In our descriptions, we adopt the nomenclature of the National Wildfire Coordinating Group (NWCG 1996, 2005) and Hardy (2005). The decision support system consists of a logic model and a decision model. In the logic model, we evaluate danger as a function of three primary topics: fire hazard, fire behavior, and risk of ignition. Each primary topic has secondary topics under which data are evaluated. The logic model shows the state of each evaluated landscape with respect to fire danger. In the decision model, we place the fire danger summary conditions of each evaluated landscape in the context of the amount of associated wildland-urban interface (WUI). The logic and decision models are executed in EMDS (Reynolds and others 2003), a decision support system that operates in ArcGIS. We show that a decision criterion such as relationship to WUI can significantly influence the outcome of a decision to determine treatment priorities. We demonstrate use of the system with an example from the Rocky Mountain region in the State of Utah, which represents a planning area of about 4.8 million ha and encompasses 575 complete subwatersheds. We discuss considerations for extending the application to support strategic planning at national and regional scales and tactical planning at local scales.

This decision support system is comparable in some aspects to the National Fire Danger Rating System (NFDRS)(Burgan 1988, Deeming and others 1977), but there are important differences and advances, too. For example, the NFDRS summarizes fire danger information pertaining to fire hazard, fire behavior, and ignition risk, the primary topics of fire danger, at a regional scale using annual weather and forest conditions information. The fire danger variables computed by FIREHARM and used in this application reflect a broader set, are computed at a stand or patch scale, and summarized to subwatersheds, and the variables are computed as probabilities of exceeding a severe fire threshold using 18 years rather than a single year of data.